src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_single/nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_sse4_1_single.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sse4_1_single kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_sse4_1_single.h"
  48 #include "kernelutil_x86_sse4_1_single.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_VF_sse4_1_single
  52  * Electrostatics interaction: Coulomb
  53  * VdW interaction:            CubicSplineTable
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_VF_sse4_1_single
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB,jnrC,jnrD;
  75     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  76     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  81     real             scratch[4*DIM];
  82     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  83     int              vdwioffset0;
  84     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  85     int              vdwioffset1;
  86     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  87     int              vdwioffset2;
  88     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  89     int              vdwioffset3;
  90     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  91     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  92     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  93     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  94     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  95     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  96     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  97     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
  98     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  99     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 100     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 101     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 102     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 103     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 104     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 105     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 106     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 107     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 108     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 109     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
 110     real             *charge;
 111     int              nvdwtype;
 112     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 113     int              *vdwtype;
 114     real             *vdwparam;
 115     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
 116     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
 117     __m128i          vfitab;
 118     __m128i          ifour       = _mm_set1_epi32(4);
 119     __m128           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 120     real             *vftab;
 121     __m128           dummy_mask,cutoff_mask;
 122     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 123     __m128           one     = _mm_set1_ps(1.0);
 124     __m128           two     = _mm_set1_ps(2.0);
 125     x                = xx[0];
 126     f                = ff[0];
 127
 128     nri              = nlist->nri;
 129     iinr             = nlist->iinr;
 130     jindex           = nlist->jindex;
 131     jjnr             = nlist->jjnr;
 132     shiftidx         = nlist->shift;
 133     gid              = nlist->gid;
 134     shiftvec         = fr->shift_vec[0];
 135     fshift           = fr->fshift[0];
 136     facel            = _mm_set1_ps(fr->epsfac);
 137     charge           = mdatoms->chargeA;
 138     nvdwtype         = fr->ntype;
 139     vdwparam         = fr->nbfp;
 140     vdwtype          = mdatoms->typeA;
 141
 142     vftab            = kernel_data->table_vdw->data;
 143     vftabscale       = _mm_set1_ps(kernel_data->table_vdw->scale);
 144
 145     /* Setup water-specific parameters */
 146     inr              = nlist->iinr[0];
 147     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
 148     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
 149     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
 150     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 151
 152     jq1              = _mm_set1_ps(charge[inr+1]);
 153     jq2              = _mm_set1_ps(charge[inr+2]);
 154     jq3              = _mm_set1_ps(charge[inr+3]);
 155     vdwjidx0A        = 2*vdwtype[inr+0];
 156     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
 157     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
 158     qq11             = _mm_mul_ps(iq1,jq1);
 159     qq12             = _mm_mul_ps(iq1,jq2);
 160     qq13             = _mm_mul_ps(iq1,jq3);
 161     qq21             = _mm_mul_ps(iq2,jq1);
 162     qq22             = _mm_mul_ps(iq2,jq2);
 163     qq23             = _mm_mul_ps(iq2,jq3);
 164     qq31             = _mm_mul_ps(iq3,jq1);
 165     qq32             = _mm_mul_ps(iq3,jq2);
 166     qq33             = _mm_mul_ps(iq3,jq3);
 167
 168     /* Avoid stupid compiler warnings */
 169     jnrA = jnrB = jnrC = jnrD = 0;
 170     j_coord_offsetA = 0;
 171     j_coord_offsetB = 0;
 172     j_coord_offsetC = 0;
 173     j_coord_offsetD = 0;
 174
 175     outeriter        = 0;
 176     inneriter        = 0;
 177
 178     for(iidx=0;iidx<4*DIM;iidx++)
 179     {
 180         scratch[iidx] = 0.0;
 181     }
 182
 183     /* Start outer loop over neighborlists */
 184     for(iidx=0; iidx<nri; iidx++)
 185     {
 186         /* Load shift vector for this list */
 187         i_shift_offset   = DIM*shiftidx[iidx];
 188
 189         /* Load limits for loop over neighbors */
 190         j_index_start    = jindex[iidx];
 191         j_index_end      = jindex[iidx+1];
 192
 193         /* Get outer coordinate index */
 194         inr              = iinr[iidx];
 195         i_coord_offset   = DIM*inr;
 196
 197         /* Load i particle coords and add shift vector */
 198         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 199                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 200
 201         fix0             = _mm_setzero_ps();
 202         fiy0             = _mm_setzero_ps();
 203         fiz0             = _mm_setzero_ps();
 204         fix1             = _mm_setzero_ps();
 205         fiy1             = _mm_setzero_ps();
 206         fiz1             = _mm_setzero_ps();
 207         fix2             = _mm_setzero_ps();
 208         fiy2             = _mm_setzero_ps();
 209         fiz2             = _mm_setzero_ps();
 210         fix3             = _mm_setzero_ps();
 211         fiy3             = _mm_setzero_ps();
 212         fiz3             = _mm_setzero_ps();
 213
 214         /* Reset potential sums */
 215         velecsum         = _mm_setzero_ps();
 216         vvdwsum          = _mm_setzero_ps();
 217
 218         /* Start inner kernel loop */
 219         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 220         {
 221
 222             /* Get j neighbor index, and coordinate index */
 223             jnrA             = jjnr[jidx];
 224             jnrB             = jjnr[jidx+1];
 225             jnrC             = jjnr[jidx+2];
 226             jnrD             = jjnr[jidx+3];
 227             j_coord_offsetA  = DIM*jnrA;
 228             j_coord_offsetB  = DIM*jnrB;
 229             j_coord_offsetC  = DIM*jnrC;
 230             j_coord_offsetD  = DIM*jnrD;
 231
 232             /* load j atom coordinates */
 233             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 234                                               x+j_coord_offsetC,x+j_coord_offsetD,
 235                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 236                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 237
 238             /* Calculate displacement vector */
 239             dx00             = _mm_sub_ps(ix0,jx0);
 240             dy00             = _mm_sub_ps(iy0,jy0);
 241             dz00             = _mm_sub_ps(iz0,jz0);
 242             dx11             = _mm_sub_ps(ix1,jx1);
 243             dy11             = _mm_sub_ps(iy1,jy1);
 244             dz11             = _mm_sub_ps(iz1,jz1);
 245             dx12             = _mm_sub_ps(ix1,jx2);
 246             dy12             = _mm_sub_ps(iy1,jy2);
 247             dz12             = _mm_sub_ps(iz1,jz2);
 248             dx13             = _mm_sub_ps(ix1,jx3);
 249             dy13             = _mm_sub_ps(iy1,jy3);
 250             dz13             = _mm_sub_ps(iz1,jz3);
 251             dx21             = _mm_sub_ps(ix2,jx1);
 252             dy21             = _mm_sub_ps(iy2,jy1);
 253             dz21             = _mm_sub_ps(iz2,jz1);
 254             dx22             = _mm_sub_ps(ix2,jx2);
 255             dy22             = _mm_sub_ps(iy2,jy2);
 256             dz22             = _mm_sub_ps(iz2,jz2);
 257             dx23             = _mm_sub_ps(ix2,jx3);
 258             dy23             = _mm_sub_ps(iy2,jy3);
 259             dz23             = _mm_sub_ps(iz2,jz3);
 260             dx31             = _mm_sub_ps(ix3,jx1);
 261             dy31             = _mm_sub_ps(iy3,jy1);
 262             dz31             = _mm_sub_ps(iz3,jz1);
 263             dx32             = _mm_sub_ps(ix3,jx2);
 264             dy32             = _mm_sub_ps(iy3,jy2);
 265             dz32             = _mm_sub_ps(iz3,jz2);
 266             dx33             = _mm_sub_ps(ix3,jx3);
 267             dy33             = _mm_sub_ps(iy3,jy3);
 268             dz33             = _mm_sub_ps(iz3,jz3);
 269
 270             /* Calculate squared distance and things based on it */
 271             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 272             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 273             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 274             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 275             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 276             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 277             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 278             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 279             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 280             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 281
 282             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 283             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 284             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 285             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 286             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 287             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 288             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 289             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 290             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 291             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 292
 293             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
 294             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
 295             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
 296             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
 297             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
 298             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
 299             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
 300             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
 301             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
 302
 303             fjx0             = _mm_setzero_ps();
 304             fjy0             = _mm_setzero_ps();
 305             fjz0             = _mm_setzero_ps();
 306             fjx1             = _mm_setzero_ps();
 307             fjy1             = _mm_setzero_ps();
 308             fjz1             = _mm_setzero_ps();
 309             fjx2             = _mm_setzero_ps();
 310             fjy2             = _mm_setzero_ps();
 311             fjz2             = _mm_setzero_ps();
 312             fjx3             = _mm_setzero_ps();
 313             fjy3             = _mm_setzero_ps();
 314             fjz3             = _mm_setzero_ps();
 315
 316             /**************************
 317              * CALCULATE INTERACTIONS *
 318              **************************/
 319
 320             r00              = _mm_mul_ps(rsq00,rinv00);
 321
 322             /* Calculate table index by multiplying r with table scale and truncate to integer */
 323             rt               = _mm_mul_ps(r00,vftabscale);
 324             vfitab           = _mm_cvttps_epi32(rt);
 325             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 326             vfitab           = _mm_slli_epi32(vfitab,3);
 327
 328             /* CUBIC SPLINE TABLE DISPERSION */
 329             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 330             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 331             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 332             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 333             _MM_TRANSPOSE4_PS(Y,F,G,H);
 334             Heps             = _mm_mul_ps(vfeps,H);
 335             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 336             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 337             vvdw6            = _mm_mul_ps(c6_00,VV);
 338             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 339             fvdw6            = _mm_mul_ps(c6_00,FF);
 340
 341             /* CUBIC SPLINE TABLE REPULSION */
 342             vfitab           = _mm_add_epi32(vfitab,ifour);
 343             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 344             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 345             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 346             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 347             _MM_TRANSPOSE4_PS(Y,F,G,H);
 348             Heps             = _mm_mul_ps(vfeps,H);
 349             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 350             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 351             vvdw12           = _mm_mul_ps(c12_00,VV);
 352             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 353             fvdw12           = _mm_mul_ps(c12_00,FF);
 354             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 355             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 356
 357             /* Update potential sum for this i atom from the interaction with this j atom. */
 358             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 359
 360             fscal            = fvdw;
 361
 362             /* Calculate temporary vectorial force */
 363             tx               = _mm_mul_ps(fscal,dx00);
 364             ty               = _mm_mul_ps(fscal,dy00);
 365             tz               = _mm_mul_ps(fscal,dz00);
 366
 367             /* Update vectorial force */
 368             fix0             = _mm_add_ps(fix0,tx);
 369             fiy0             = _mm_add_ps(fiy0,ty);
 370             fiz0             = _mm_add_ps(fiz0,tz);
 371
 372             fjx0             = _mm_add_ps(fjx0,tx);
 373             fjy0             = _mm_add_ps(fjy0,ty);
 374             fjz0             = _mm_add_ps(fjz0,tz);
 375
 376             /**************************
 377              * CALCULATE INTERACTIONS *
 378              **************************/
 379
 380             /* COULOMB ELECTROSTATICS */
 381             velec            = _mm_mul_ps(qq11,rinv11);
 382             felec            = _mm_mul_ps(velec,rinvsq11);
 383
 384             /* Update potential sum for this i atom from the interaction with this j atom. */
 385             velecsum         = _mm_add_ps(velecsum,velec);
 386
 387             fscal            = felec;
 388
 389             /* Calculate temporary vectorial force */
 390             tx               = _mm_mul_ps(fscal,dx11);
 391             ty               = _mm_mul_ps(fscal,dy11);
 392             tz               = _mm_mul_ps(fscal,dz11);
 393
 394             /* Update vectorial force */
 395             fix1             = _mm_add_ps(fix1,tx);
 396             fiy1             = _mm_add_ps(fiy1,ty);
 397             fiz1             = _mm_add_ps(fiz1,tz);
 398
 399             fjx1             = _mm_add_ps(fjx1,tx);
 400             fjy1             = _mm_add_ps(fjy1,ty);
 401             fjz1             = _mm_add_ps(fjz1,tz);
 402
 403             /**************************
 404              * CALCULATE INTERACTIONS *
 405              **************************/
 406
 407             /* COULOMB ELECTROSTATICS */
 408             velec            = _mm_mul_ps(qq12,rinv12);
 409             felec            = _mm_mul_ps(velec,rinvsq12);
 410
 411             /* Update potential sum for this i atom from the interaction with this j atom. */
 412             velecsum         = _mm_add_ps(velecsum,velec);
 413
 414             fscal            = felec;
 415
 416             /* Calculate temporary vectorial force */
 417             tx               = _mm_mul_ps(fscal,dx12);
 418             ty               = _mm_mul_ps(fscal,dy12);
 419             tz               = _mm_mul_ps(fscal,dz12);
 420
 421             /* Update vectorial force */
 422             fix1             = _mm_add_ps(fix1,tx);
 423             fiy1             = _mm_add_ps(fiy1,ty);
 424             fiz1             = _mm_add_ps(fiz1,tz);
 425
 426             fjx2             = _mm_add_ps(fjx2,tx);
 427             fjy2             = _mm_add_ps(fjy2,ty);
 428             fjz2             = _mm_add_ps(fjz2,tz);
 429
 430             /**************************
 431              * CALCULATE INTERACTIONS *
 432              **************************/
 433
 434             /* COULOMB ELECTROSTATICS */
 435             velec            = _mm_mul_ps(qq13,rinv13);
 436             felec            = _mm_mul_ps(velec,rinvsq13);
 437
 438             /* Update potential sum for this i atom from the interaction with this j atom. */
 439             velecsum         = _mm_add_ps(velecsum,velec);
 440
 441             fscal            = felec;
 442
 443             /* Calculate temporary vectorial force */
 444             tx               = _mm_mul_ps(fscal,dx13);
 445             ty               = _mm_mul_ps(fscal,dy13);
 446             tz               = _mm_mul_ps(fscal,dz13);
 447
 448             /* Update vectorial force */
 449             fix1             = _mm_add_ps(fix1,tx);
 450             fiy1             = _mm_add_ps(fiy1,ty);
 451             fiz1             = _mm_add_ps(fiz1,tz);
 452
 453             fjx3             = _mm_add_ps(fjx3,tx);
 454             fjy3             = _mm_add_ps(fjy3,ty);
 455             fjz3             = _mm_add_ps(fjz3,tz);
 456
 457             /**************************
 458              * CALCULATE INTERACTIONS *
 459              **************************/
 460
 461             /* COULOMB ELECTROSTATICS */
 462             velec            = _mm_mul_ps(qq21,rinv21);
 463             felec            = _mm_mul_ps(velec,rinvsq21);
 464
 465             /* Update potential sum for this i atom from the interaction with this j atom. */
 466             velecsum         = _mm_add_ps(velecsum,velec);
 467
 468             fscal            = felec;
 469
 470             /* Calculate temporary vectorial force */
 471             tx               = _mm_mul_ps(fscal,dx21);
 472             ty               = _mm_mul_ps(fscal,dy21);
 473             tz               = _mm_mul_ps(fscal,dz21);
 474
 475             /* Update vectorial force */
 476             fix2             = _mm_add_ps(fix2,tx);
 477             fiy2             = _mm_add_ps(fiy2,ty);
 478             fiz2             = _mm_add_ps(fiz2,tz);
 479
 480             fjx1             = _mm_add_ps(fjx1,tx);
 481             fjy1             = _mm_add_ps(fjy1,ty);
 482             fjz1             = _mm_add_ps(fjz1,tz);
 483
 484             /**************************
 485              * CALCULATE INTERACTIONS *
 486              **************************/
 487
 488             /* COULOMB ELECTROSTATICS */
 489             velec            = _mm_mul_ps(qq22,rinv22);
 490             felec            = _mm_mul_ps(velec,rinvsq22);
 491
 492             /* Update potential sum for this i atom from the interaction with this j atom. */
 493             velecsum         = _mm_add_ps(velecsum,velec);
 494
 495             fscal            = felec;
 496
 497             /* Calculate temporary vectorial force */
 498             tx               = _mm_mul_ps(fscal,dx22);
 499             ty               = _mm_mul_ps(fscal,dy22);
 500             tz               = _mm_mul_ps(fscal,dz22);
 501
 502             /* Update vectorial force */
 503             fix2             = _mm_add_ps(fix2,tx);
 504             fiy2             = _mm_add_ps(fiy2,ty);
 505             fiz2             = _mm_add_ps(fiz2,tz);
 506
 507             fjx2             = _mm_add_ps(fjx2,tx);
 508             fjy2             = _mm_add_ps(fjy2,ty);
 509             fjz2             = _mm_add_ps(fjz2,tz);
 510
 511             /**************************
 512              * CALCULATE INTERACTIONS *
 513              **************************/
 514
 515             /* COULOMB ELECTROSTATICS */
 516             velec            = _mm_mul_ps(qq23,rinv23);
 517             felec            = _mm_mul_ps(velec,rinvsq23);
 518
 519             /* Update potential sum for this i atom from the interaction with this j atom. */
 520             velecsum         = _mm_add_ps(velecsum,velec);
 521
 522             fscal            = felec;
 523
 524             /* Calculate temporary vectorial force */
 525             tx               = _mm_mul_ps(fscal,dx23);
 526             ty               = _mm_mul_ps(fscal,dy23);
 527             tz               = _mm_mul_ps(fscal,dz23);
 528
 529             /* Update vectorial force */
 530             fix2             = _mm_add_ps(fix2,tx);
 531             fiy2             = _mm_add_ps(fiy2,ty);
 532             fiz2             = _mm_add_ps(fiz2,tz);
 533
 534             fjx3             = _mm_add_ps(fjx3,tx);
 535             fjy3             = _mm_add_ps(fjy3,ty);
 536             fjz3             = _mm_add_ps(fjz3,tz);
 537
 538             /**************************
 539              * CALCULATE INTERACTIONS *
 540              **************************/
 541
 542             /* COULOMB ELECTROSTATICS */
 543             velec            = _mm_mul_ps(qq31,rinv31);
 544             felec            = _mm_mul_ps(velec,rinvsq31);
 545
 546             /* Update potential sum for this i atom from the interaction with this j atom. */
 547             velecsum         = _mm_add_ps(velecsum,velec);
 548
 549             fscal            = felec;
 550
 551             /* Calculate temporary vectorial force */
 552             tx               = _mm_mul_ps(fscal,dx31);
 553             ty               = _mm_mul_ps(fscal,dy31);
 554             tz               = _mm_mul_ps(fscal,dz31);
 555
 556             /* Update vectorial force */
 557             fix3             = _mm_add_ps(fix3,tx);
 558             fiy3             = _mm_add_ps(fiy3,ty);
 559             fiz3             = _mm_add_ps(fiz3,tz);
 560
 561             fjx1             = _mm_add_ps(fjx1,tx);
 562             fjy1             = _mm_add_ps(fjy1,ty);
 563             fjz1             = _mm_add_ps(fjz1,tz);
 564
 565             /**************************
 566              * CALCULATE INTERACTIONS *
 567              **************************/
 568
 569             /* COULOMB ELECTROSTATICS */
 570             velec            = _mm_mul_ps(qq32,rinv32);
 571             felec            = _mm_mul_ps(velec,rinvsq32);
 572
 573             /* Update potential sum for this i atom from the interaction with this j atom. */
 574             velecsum         = _mm_add_ps(velecsum,velec);
 575
 576             fscal            = felec;
 577
 578             /* Calculate temporary vectorial force */
 579             tx               = _mm_mul_ps(fscal,dx32);
 580             ty               = _mm_mul_ps(fscal,dy32);
 581             tz               = _mm_mul_ps(fscal,dz32);
 582
 583             /* Update vectorial force */
 584             fix3             = _mm_add_ps(fix3,tx);
 585             fiy3             = _mm_add_ps(fiy3,ty);
 586             fiz3             = _mm_add_ps(fiz3,tz);
 587
 588             fjx2             = _mm_add_ps(fjx2,tx);
 589             fjy2             = _mm_add_ps(fjy2,ty);
 590             fjz2             = _mm_add_ps(fjz2,tz);
 591
 592             /**************************
 593              * CALCULATE INTERACTIONS *
 594              **************************/
 595
 596             /* COULOMB ELECTROSTATICS */
 597             velec            = _mm_mul_ps(qq33,rinv33);
 598             felec            = _mm_mul_ps(velec,rinvsq33);
 599
 600             /* Update potential sum for this i atom from the interaction with this j atom. */
 601             velecsum         = _mm_add_ps(velecsum,velec);
 602
 603             fscal            = felec;
 604
 605             /* Calculate temporary vectorial force */
 606             tx               = _mm_mul_ps(fscal,dx33);
 607             ty               = _mm_mul_ps(fscal,dy33);
 608             tz               = _mm_mul_ps(fscal,dz33);
 609
 610             /* Update vectorial force */
 611             fix3             = _mm_add_ps(fix3,tx);
 612             fiy3             = _mm_add_ps(fiy3,ty);
 613             fiz3             = _mm_add_ps(fiz3,tz);
 614
 615             fjx3             = _mm_add_ps(fjx3,tx);
 616             fjy3             = _mm_add_ps(fjy3,ty);
 617             fjz3             = _mm_add_ps(fjz3,tz);
 618
 619             fjptrA             = f+j_coord_offsetA;
 620             fjptrB             = f+j_coord_offsetB;
 621             fjptrC             = f+j_coord_offsetC;
 622             fjptrD             = f+j_coord_offsetD;
 623
 624             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 625                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 626                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 627
 628             /* Inner loop uses 311 flops */
 629         }
 630
 631         if(jidx<j_index_end)
 632         {
 633
 634             /* Get j neighbor index, and coordinate index */
 635             jnrlistA         = jjnr[jidx];
 636             jnrlistB         = jjnr[jidx+1];
 637             jnrlistC         = jjnr[jidx+2];
 638             jnrlistD         = jjnr[jidx+3];
 639             /* Sign of each element will be negative for non-real atoms.
 640              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 641              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 642              */
 643             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 644             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 645             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 646             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 647             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 648             j_coord_offsetA  = DIM*jnrA;
 649             j_coord_offsetB  = DIM*jnrB;
 650             j_coord_offsetC  = DIM*jnrC;
 651             j_coord_offsetD  = DIM*jnrD;
 652
 653             /* load j atom coordinates */
 654             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 655                                               x+j_coord_offsetC,x+j_coord_offsetD,
 656                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 657                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 658
 659             /* Calculate displacement vector */
 660             dx00             = _mm_sub_ps(ix0,jx0);
 661             dy00             = _mm_sub_ps(iy0,jy0);
 662             dz00             = _mm_sub_ps(iz0,jz0);
 663             dx11             = _mm_sub_ps(ix1,jx1);
 664             dy11             = _mm_sub_ps(iy1,jy1);
 665             dz11             = _mm_sub_ps(iz1,jz1);
 666             dx12             = _mm_sub_ps(ix1,jx2);
 667             dy12             = _mm_sub_ps(iy1,jy2);
 668             dz12             = _mm_sub_ps(iz1,jz2);
 669             dx13             = _mm_sub_ps(ix1,jx3);
 670             dy13             = _mm_sub_ps(iy1,jy3);
 671             dz13             = _mm_sub_ps(iz1,jz3);
 672             dx21             = _mm_sub_ps(ix2,jx1);
 673             dy21             = _mm_sub_ps(iy2,jy1);
 674             dz21             = _mm_sub_ps(iz2,jz1);
 675             dx22             = _mm_sub_ps(ix2,jx2);
 676             dy22             = _mm_sub_ps(iy2,jy2);
 677             dz22             = _mm_sub_ps(iz2,jz2);
 678             dx23             = _mm_sub_ps(ix2,jx3);
 679             dy23             = _mm_sub_ps(iy2,jy3);
 680             dz23             = _mm_sub_ps(iz2,jz3);
 681             dx31             = _mm_sub_ps(ix3,jx1);
 682             dy31             = _mm_sub_ps(iy3,jy1);
 683             dz31             = _mm_sub_ps(iz3,jz1);
 684             dx32             = _mm_sub_ps(ix3,jx2);
 685             dy32             = _mm_sub_ps(iy3,jy2);
 686             dz32             = _mm_sub_ps(iz3,jz2);
 687             dx33             = _mm_sub_ps(ix3,jx3);
 688             dy33             = _mm_sub_ps(iy3,jy3);
 689             dz33             = _mm_sub_ps(iz3,jz3);
 690
 691             /* Calculate squared distance and things based on it */
 692             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 693             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 694             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 695             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 696             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 697             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 698             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 699             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 700             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 701             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 702
 703             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 704             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 705             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 706             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 707             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 708             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 709             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 710             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 711             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 712             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 713
 714             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
 715             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
 716             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
 717             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
 718             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
 719             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
 720             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
 721             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
 722             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
 723
 724             fjx0             = _mm_setzero_ps();
 725             fjy0             = _mm_setzero_ps();
 726             fjz0             = _mm_setzero_ps();
 727             fjx1             = _mm_setzero_ps();
 728             fjy1             = _mm_setzero_ps();
 729             fjz1             = _mm_setzero_ps();
 730             fjx2             = _mm_setzero_ps();
 731             fjy2             = _mm_setzero_ps();
 732             fjz2             = _mm_setzero_ps();
 733             fjx3             = _mm_setzero_ps();
 734             fjy3             = _mm_setzero_ps();
 735             fjz3             = _mm_setzero_ps();
 736
 737             /**************************
 738              * CALCULATE INTERACTIONS *
 739              **************************/
 740
 741             r00              = _mm_mul_ps(rsq00,rinv00);
 742             r00              = _mm_andnot_ps(dummy_mask,r00);
 743
 744             /* Calculate table index by multiplying r with table scale and truncate to integer */
 745             rt               = _mm_mul_ps(r00,vftabscale);
 746             vfitab           = _mm_cvttps_epi32(rt);
 747             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 748             vfitab           = _mm_slli_epi32(vfitab,3);
 749
 750             /* CUBIC SPLINE TABLE DISPERSION */
 751             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 752             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 753             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 754             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 755             _MM_TRANSPOSE4_PS(Y,F,G,H);
 756             Heps             = _mm_mul_ps(vfeps,H);
 757             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 758             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 759             vvdw6            = _mm_mul_ps(c6_00,VV);
 760             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 761             fvdw6            = _mm_mul_ps(c6_00,FF);
 762
 763             /* CUBIC SPLINE TABLE REPULSION */
 764             vfitab           = _mm_add_epi32(vfitab,ifour);
 765             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 766             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 767             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 768             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 769             _MM_TRANSPOSE4_PS(Y,F,G,H);
 770             Heps             = _mm_mul_ps(vfeps,H);
 771             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 772             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 773             vvdw12           = _mm_mul_ps(c12_00,VV);
 774             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 775             fvdw12           = _mm_mul_ps(c12_00,FF);
 776             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 777             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 778
 779             /* Update potential sum for this i atom from the interaction with this j atom. */
 780             vvdw             = _mm_andnot_ps(dummy_mask,vvdw);
 781             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 782
 783             fscal            = fvdw;
 784
 785             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 786
 787             /* Calculate temporary vectorial force */
 788             tx               = _mm_mul_ps(fscal,dx00);
 789             ty               = _mm_mul_ps(fscal,dy00);
 790             tz               = _mm_mul_ps(fscal,dz00);
 791
 792             /* Update vectorial force */
 793             fix0             = _mm_add_ps(fix0,tx);
 794             fiy0             = _mm_add_ps(fiy0,ty);
 795             fiz0             = _mm_add_ps(fiz0,tz);
 796
 797             fjx0             = _mm_add_ps(fjx0,tx);
 798             fjy0             = _mm_add_ps(fjy0,ty);
 799             fjz0             = _mm_add_ps(fjz0,tz);
 800
 801             /**************************
 802              * CALCULATE INTERACTIONS *
 803              **************************/
 804
 805             /* COULOMB ELECTROSTATICS */
 806             velec            = _mm_mul_ps(qq11,rinv11);
 807             felec            = _mm_mul_ps(velec,rinvsq11);
 808
 809             /* Update potential sum for this i atom from the interaction with this j atom. */
 810             velec            = _mm_andnot_ps(dummy_mask,velec);
 811             velecsum         = _mm_add_ps(velecsum,velec);
 812
 813             fscal            = felec;
 814
 815             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 816
 817             /* Calculate temporary vectorial force */
 818             tx               = _mm_mul_ps(fscal,dx11);
 819             ty               = _mm_mul_ps(fscal,dy11);
 820             tz               = _mm_mul_ps(fscal,dz11);
 821
 822             /* Update vectorial force */
 823             fix1             = _mm_add_ps(fix1,tx);
 824             fiy1             = _mm_add_ps(fiy1,ty);
 825             fiz1             = _mm_add_ps(fiz1,tz);
 826
 827             fjx1             = _mm_add_ps(fjx1,tx);
 828             fjy1             = _mm_add_ps(fjy1,ty);
 829             fjz1             = _mm_add_ps(fjz1,tz);
 830
 831             /**************************
 832              * CALCULATE INTERACTIONS *
 833              **************************/
 834
 835             /* COULOMB ELECTROSTATICS */
 836             velec            = _mm_mul_ps(qq12,rinv12);
 837             felec            = _mm_mul_ps(velec,rinvsq12);
 838
 839             /* Update potential sum for this i atom from the interaction with this j atom. */
 840             velec            = _mm_andnot_ps(dummy_mask,velec);
 841             velecsum         = _mm_add_ps(velecsum,velec);
 842
 843             fscal            = felec;
 844
 845             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 846
 847             /* Calculate temporary vectorial force */
 848             tx               = _mm_mul_ps(fscal,dx12);
 849             ty               = _mm_mul_ps(fscal,dy12);
 850             tz               = _mm_mul_ps(fscal,dz12);
 851
 852             /* Update vectorial force */
 853             fix1             = _mm_add_ps(fix1,tx);
 854             fiy1             = _mm_add_ps(fiy1,ty);
 855             fiz1             = _mm_add_ps(fiz1,tz);
 856
 857             fjx2             = _mm_add_ps(fjx2,tx);
 858             fjy2             = _mm_add_ps(fjy2,ty);
 859             fjz2             = _mm_add_ps(fjz2,tz);
 860
 861             /**************************
 862              * CALCULATE INTERACTIONS *
 863              **************************/
 864
 865             /* COULOMB ELECTROSTATICS */
 866             velec            = _mm_mul_ps(qq13,rinv13);
 867             felec            = _mm_mul_ps(velec,rinvsq13);
 868
 869             /* Update potential sum for this i atom from the interaction with this j atom. */
 870             velec            = _mm_andnot_ps(dummy_mask,velec);
 871             velecsum         = _mm_add_ps(velecsum,velec);
 872
 873             fscal            = felec;
 874
 875             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 876
 877             /* Calculate temporary vectorial force */
 878             tx               = _mm_mul_ps(fscal,dx13);
 879             ty               = _mm_mul_ps(fscal,dy13);
 880             tz               = _mm_mul_ps(fscal,dz13);
 881
 882             /* Update vectorial force */
 883             fix1             = _mm_add_ps(fix1,tx);
 884             fiy1             = _mm_add_ps(fiy1,ty);
 885             fiz1             = _mm_add_ps(fiz1,tz);
 886
 887             fjx3             = _mm_add_ps(fjx3,tx);
 888             fjy3             = _mm_add_ps(fjy3,ty);
 889             fjz3             = _mm_add_ps(fjz3,tz);
 890
 891             /**************************
 892              * CALCULATE INTERACTIONS *
 893              **************************/
 894
 895             /* COULOMB ELECTROSTATICS */
 896             velec            = _mm_mul_ps(qq21,rinv21);
 897             felec            = _mm_mul_ps(velec,rinvsq21);
 898
 899             /* Update potential sum for this i atom from the interaction with this j atom. */
 900             velec            = _mm_andnot_ps(dummy_mask,velec);
 901             velecsum         = _mm_add_ps(velecsum,velec);
 902
 903             fscal            = felec;
 904
 905             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 906
 907             /* Calculate temporary vectorial force */
 908             tx               = _mm_mul_ps(fscal,dx21);
 909             ty               = _mm_mul_ps(fscal,dy21);
 910             tz               = _mm_mul_ps(fscal,dz21);
 911
 912             /* Update vectorial force */
 913             fix2             = _mm_add_ps(fix2,tx);
 914             fiy2             = _mm_add_ps(fiy2,ty);
 915             fiz2             = _mm_add_ps(fiz2,tz);
 916
 917             fjx1             = _mm_add_ps(fjx1,tx);
 918             fjy1             = _mm_add_ps(fjy1,ty);
 919             fjz1             = _mm_add_ps(fjz1,tz);
 920
 921             /**************************
 922              * CALCULATE INTERACTIONS *
 923              **************************/
 924
 925             /* COULOMB ELECTROSTATICS */
 926             velec            = _mm_mul_ps(qq22,rinv22);
 927             felec            = _mm_mul_ps(velec,rinvsq22);
 928
 929             /* Update potential sum for this i atom from the interaction with this j atom. */
 930             velec            = _mm_andnot_ps(dummy_mask,velec);
 931             velecsum         = _mm_add_ps(velecsum,velec);
 932
 933             fscal            = felec;
 934
 935             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 936
 937             /* Calculate temporary vectorial force */
 938             tx               = _mm_mul_ps(fscal,dx22);
 939             ty               = _mm_mul_ps(fscal,dy22);
 940             tz               = _mm_mul_ps(fscal,dz22);
 941
 942             /* Update vectorial force */
 943             fix2             = _mm_add_ps(fix2,tx);
 944             fiy2             = _mm_add_ps(fiy2,ty);
 945             fiz2             = _mm_add_ps(fiz2,tz);
 946
 947             fjx2             = _mm_add_ps(fjx2,tx);
 948             fjy2             = _mm_add_ps(fjy2,ty);
 949             fjz2             = _mm_add_ps(fjz2,tz);
 950
 951             /**************************
 952              * CALCULATE INTERACTIONS *
 953              **************************/
 954
 955             /* COULOMB ELECTROSTATICS */
 956             velec            = _mm_mul_ps(qq23,rinv23);
 957             felec            = _mm_mul_ps(velec,rinvsq23);
 958
 959             /* Update potential sum for this i atom from the interaction with this j atom. */
 960             velec            = _mm_andnot_ps(dummy_mask,velec);
 961             velecsum         = _mm_add_ps(velecsum,velec);
 962
 963             fscal            = felec;
 964
 965             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 966
 967             /* Calculate temporary vectorial force */
 968             tx               = _mm_mul_ps(fscal,dx23);
 969             ty               = _mm_mul_ps(fscal,dy23);
 970             tz               = _mm_mul_ps(fscal,dz23);
 971
 972             /* Update vectorial force */
 973             fix2             = _mm_add_ps(fix2,tx);
 974             fiy2             = _mm_add_ps(fiy2,ty);
 975             fiz2             = _mm_add_ps(fiz2,tz);
 976
 977             fjx3             = _mm_add_ps(fjx3,tx);
 978             fjy3             = _mm_add_ps(fjy3,ty);
 979             fjz3             = _mm_add_ps(fjz3,tz);
 980
 981             /**************************
 982              * CALCULATE INTERACTIONS *
 983              **************************/
 984
 985             /* COULOMB ELECTROSTATICS */
 986             velec            = _mm_mul_ps(qq31,rinv31);
 987             felec            = _mm_mul_ps(velec,rinvsq31);
 988
 989             /* Update potential sum for this i atom from the interaction with this j atom. */
 990             velec            = _mm_andnot_ps(dummy_mask,velec);
 991             velecsum         = _mm_add_ps(velecsum,velec);
 992
 993             fscal            = felec;
 994
 995             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 996
 997             /* Calculate temporary vectorial force */
 998             tx               = _mm_mul_ps(fscal,dx31);
 999             ty               = _mm_mul_ps(fscal,dy31);
1000             tz               = _mm_mul_ps(fscal,dz31);
1001
1002             /* Update vectorial force */
1003             fix3             = _mm_add_ps(fix3,tx);
1004             fiy3             = _mm_add_ps(fiy3,ty);
1005             fiz3             = _mm_add_ps(fiz3,tz);
1006
1007             fjx1             = _mm_add_ps(fjx1,tx);
1008             fjy1             = _mm_add_ps(fjy1,ty);
1009             fjz1             = _mm_add_ps(fjz1,tz);
1010
1011             /**************************
1012              * CALCULATE INTERACTIONS *
1013              **************************/
1014
1015             /* COULOMB ELECTROSTATICS */
1016             velec            = _mm_mul_ps(qq32,rinv32);
1017             felec            = _mm_mul_ps(velec,rinvsq32);
1018
1019             /* Update potential sum for this i atom from the interaction with this j atom. */
1020             velec            = _mm_andnot_ps(dummy_mask,velec);
1021             velecsum         = _mm_add_ps(velecsum,velec);
1022
1023             fscal            = felec;
1024
1025             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1026
1027             /* Calculate temporary vectorial force */
1028             tx               = _mm_mul_ps(fscal,dx32);
1029             ty               = _mm_mul_ps(fscal,dy32);
1030             tz               = _mm_mul_ps(fscal,dz32);
1031
1032             /* Update vectorial force */
1033             fix3             = _mm_add_ps(fix3,tx);
1034             fiy3             = _mm_add_ps(fiy3,ty);
1035             fiz3             = _mm_add_ps(fiz3,tz);
1036
1037             fjx2             = _mm_add_ps(fjx2,tx);
1038             fjy2             = _mm_add_ps(fjy2,ty);
1039             fjz2             = _mm_add_ps(fjz2,tz);
1040
1041             /**************************
1042              * CALCULATE INTERACTIONS *
1043              **************************/
1044
1045             /* COULOMB ELECTROSTATICS */
1046             velec            = _mm_mul_ps(qq33,rinv33);
1047             felec            = _mm_mul_ps(velec,rinvsq33);
1048
1049             /* Update potential sum for this i atom from the interaction with this j atom. */
1050             velec            = _mm_andnot_ps(dummy_mask,velec);
1051             velecsum         = _mm_add_ps(velecsum,velec);
1052
1053             fscal            = felec;
1054
1055             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1056
1057             /* Calculate temporary vectorial force */
1058             tx               = _mm_mul_ps(fscal,dx33);
1059             ty               = _mm_mul_ps(fscal,dy33);
1060             tz               = _mm_mul_ps(fscal,dz33);
1061
1062             /* Update vectorial force */
1063             fix3             = _mm_add_ps(fix3,tx);
1064             fiy3             = _mm_add_ps(fiy3,ty);
1065             fiz3             = _mm_add_ps(fiz3,tz);
1066
1067             fjx3             = _mm_add_ps(fjx3,tx);
1068             fjy3             = _mm_add_ps(fjy3,ty);
1069             fjz3             = _mm_add_ps(fjz3,tz);
1070
1071             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1072             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1073             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1074             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1075
1076             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1077                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1078                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1079
1080             /* Inner loop uses 312 flops */
1081         }
1082
1083         /* End of innermost loop */
1084
1085         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1086                                               f+i_coord_offset,fshift+i_shift_offset);
1087
1088         ggid                        = gid[iidx];
1089         /* Update potential energies */
1090         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1091         gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1092
1093         /* Increment number of inner iterations */
1094         inneriter                  += j_index_end - j_index_start;
1095
1096         /* Outer loop uses 26 flops */
1097     }
1098
1099     /* Increment number of outer iterations */
1100     outeriter        += nri;
1101
1102     /* Update outer/inner flops */
1103
1104     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*312);
1105 }
1106 /*
1107  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_F_sse4_1_single
1108  * Electrostatics interaction: Coulomb
1109  * VdW interaction:            CubicSplineTable
1110  * Geometry:                   Water4-Water4
1111  * Calculate force/pot:        Force
1112  */
1113 void
1114 nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_F_sse4_1_single
1115                     (t_nblist                    * gmx_restrict       nlist,
1116                      rvec                        * gmx_restrict          xx,
1117                      rvec                        * gmx_restrict          ff,
1118                      t_forcerec                  * gmx_restrict          fr,
1119                      t_mdatoms                   * gmx_restrict     mdatoms,
1120                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1121                      t_nrnb                      * gmx_restrict        nrnb)
1122 {
1123     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1124      * just 0 for non-waters.
1125      * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
1126      * jnr indices corresponding to data put in the four positions in the SIMD register.
1127      */
1128     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1129     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1130     int              jnrA,jnrB,jnrC,jnrD;
1131     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1132     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1133     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1134     real             rcutoff_scalar;
1135     real             *shiftvec,*fshift,*x,*f;
1136     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1137     real             scratch[4*DIM];
1138     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1139     int              vdwioffset0;
1140     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1141     int              vdwioffset1;
1142     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1143     int              vdwioffset2;
1144     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1145     int              vdwioffset3;
1146     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1147     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1148     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1149     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1150     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1151     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1152     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1153     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1154     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1155     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1156     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1157     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1158     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1159     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1160     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1161     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1162     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1163     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1164     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1165     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
1166     real             *charge;
1167     int              nvdwtype;
1168     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1169     int              *vdwtype;
1170     real             *vdwparam;
1171     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
1172     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
1173     __m128i          vfitab;
1174     __m128i          ifour       = _mm_set1_epi32(4);
1175     __m128           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1176     real             *vftab;
1177     __m128           dummy_mask,cutoff_mask;
1178     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1179     __m128           one     = _mm_set1_ps(1.0);
1180     __m128           two     = _mm_set1_ps(2.0);
1181     x                = xx[0];
1182     f                = ff[0];
1183
1184     nri              = nlist->nri;
1185     iinr             = nlist->iinr;
1186     jindex           = nlist->jindex;
1187     jjnr             = nlist->jjnr;
1188     shiftidx         = nlist->shift;
1189     gid              = nlist->gid;
1190     shiftvec         = fr->shift_vec[0];
1191     fshift           = fr->fshift[0];
1192     facel            = _mm_set1_ps(fr->epsfac);
1193     charge           = mdatoms->chargeA;
1194     nvdwtype         = fr->ntype;
1195     vdwparam         = fr->nbfp;
1196     vdwtype          = mdatoms->typeA;
1197
1198     vftab            = kernel_data->table_vdw->data;
1199     vftabscale       = _mm_set1_ps(kernel_data->table_vdw->scale);
1200
1201     /* Setup water-specific parameters */
1202     inr              = nlist->iinr[0];
1203     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1204     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1205     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
1206     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1207
1208     jq1              = _mm_set1_ps(charge[inr+1]);
1209     jq2              = _mm_set1_ps(charge[inr+2]);
1210     jq3              = _mm_set1_ps(charge[inr+3]);
1211     vdwjidx0A        = 2*vdwtype[inr+0];
1212     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
1213     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
1214     qq11             = _mm_mul_ps(iq1,jq1);
1215     qq12             = _mm_mul_ps(iq1,jq2);
1216     qq13             = _mm_mul_ps(iq1,jq3);
1217     qq21             = _mm_mul_ps(iq2,jq1);
1218     qq22             = _mm_mul_ps(iq2,jq2);
1219     qq23             = _mm_mul_ps(iq2,jq3);
1220     qq31             = _mm_mul_ps(iq3,jq1);
1221     qq32             = _mm_mul_ps(iq3,jq2);
1222     qq33             = _mm_mul_ps(iq3,jq3);
1223
1224     /* Avoid stupid compiler warnings */
1225     jnrA = jnrB = jnrC = jnrD = 0;
1226     j_coord_offsetA = 0;
1227     j_coord_offsetB = 0;
1228     j_coord_offsetC = 0;
1229     j_coord_offsetD = 0;
1230
1231     outeriter        = 0;
1232     inneriter        = 0;
1233
1234     for(iidx=0;iidx<4*DIM;iidx++)
1235     {
1236         scratch[iidx] = 0.0;
1237     }
1238
1239     /* Start outer loop over neighborlists */
1240     for(iidx=0; iidx<nri; iidx++)
1241     {
1242         /* Load shift vector for this list */
1243         i_shift_offset   = DIM*shiftidx[iidx];
1244
1245         /* Load limits for loop over neighbors */
1246         j_index_start    = jindex[iidx];
1247         j_index_end      = jindex[iidx+1];
1248
1249         /* Get outer coordinate index */
1250         inr              = iinr[iidx];
1251         i_coord_offset   = DIM*inr;
1252
1253         /* Load i particle coords and add shift vector */
1254         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1255                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1256
1257         fix0             = _mm_setzero_ps();
1258         fiy0             = _mm_setzero_ps();
1259         fiz0             = _mm_setzero_ps();
1260         fix1             = _mm_setzero_ps();
1261         fiy1             = _mm_setzero_ps();
1262         fiz1             = _mm_setzero_ps();
1263         fix2             = _mm_setzero_ps();
1264         fiy2             = _mm_setzero_ps();
1265         fiz2             = _mm_setzero_ps();
1266         fix3             = _mm_setzero_ps();
1267         fiy3             = _mm_setzero_ps();
1268         fiz3             = _mm_setzero_ps();
1269
1270         /* Start inner kernel loop */
1271         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1272         {
1273
1274             /* Get j neighbor index, and coordinate index */
1275             jnrA             = jjnr[jidx];
1276             jnrB             = jjnr[jidx+1];
1277             jnrC             = jjnr[jidx+2];
1278             jnrD             = jjnr[jidx+3];
1279             j_coord_offsetA  = DIM*jnrA;
1280             j_coord_offsetB  = DIM*jnrB;
1281             j_coord_offsetC  = DIM*jnrC;
1282             j_coord_offsetD  = DIM*jnrD;
1283
1284             /* load j atom coordinates */
1285             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1286                                               x+j_coord_offsetC,x+j_coord_offsetD,
1287                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1288                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1289
1290             /* Calculate displacement vector */
1291             dx00             = _mm_sub_ps(ix0,jx0);
1292             dy00             = _mm_sub_ps(iy0,jy0);
1293             dz00             = _mm_sub_ps(iz0,jz0);
1294             dx11             = _mm_sub_ps(ix1,jx1);
1295             dy11             = _mm_sub_ps(iy1,jy1);
1296             dz11             = _mm_sub_ps(iz1,jz1);
1297             dx12             = _mm_sub_ps(ix1,jx2);
1298             dy12             = _mm_sub_ps(iy1,jy2);
1299             dz12             = _mm_sub_ps(iz1,jz2);
1300             dx13             = _mm_sub_ps(ix1,jx3);
1301             dy13             = _mm_sub_ps(iy1,jy3);
1302             dz13             = _mm_sub_ps(iz1,jz3);
1303             dx21             = _mm_sub_ps(ix2,jx1);
1304             dy21             = _mm_sub_ps(iy2,jy1);
1305             dz21             = _mm_sub_ps(iz2,jz1);
1306             dx22             = _mm_sub_ps(ix2,jx2);
1307             dy22             = _mm_sub_ps(iy2,jy2);
1308             dz22             = _mm_sub_ps(iz2,jz2);
1309             dx23             = _mm_sub_ps(ix2,jx3);
1310             dy23             = _mm_sub_ps(iy2,jy3);
1311             dz23             = _mm_sub_ps(iz2,jz3);
1312             dx31             = _mm_sub_ps(ix3,jx1);
1313             dy31             = _mm_sub_ps(iy3,jy1);
1314             dz31             = _mm_sub_ps(iz3,jz1);
1315             dx32             = _mm_sub_ps(ix3,jx2);
1316             dy32             = _mm_sub_ps(iy3,jy2);
1317             dz32             = _mm_sub_ps(iz3,jz2);
1318             dx33             = _mm_sub_ps(ix3,jx3);
1319             dy33             = _mm_sub_ps(iy3,jy3);
1320             dz33             = _mm_sub_ps(iz3,jz3);
1321
1322             /* Calculate squared distance and things based on it */
1323             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1324             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1325             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1326             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1327             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1328             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1329             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1330             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1331             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1332             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1333
1334             rinv00           = gmx_mm_invsqrt_ps(rsq00);
1335             rinv11           = gmx_mm_invsqrt_ps(rsq11);
1336             rinv12           = gmx_mm_invsqrt_ps(rsq12);
1337             rinv13           = gmx_mm_invsqrt_ps(rsq13);
1338             rinv21           = gmx_mm_invsqrt_ps(rsq21);
1339             rinv22           = gmx_mm_invsqrt_ps(rsq22);
1340             rinv23           = gmx_mm_invsqrt_ps(rsq23);
1341             rinv31           = gmx_mm_invsqrt_ps(rsq31);
1342             rinv32           = gmx_mm_invsqrt_ps(rsq32);
1343             rinv33           = gmx_mm_invsqrt_ps(rsq33);
1344
1345             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
1346             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
1347             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
1348             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
1349             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
1350             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
1351             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
1352             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
1353             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
1354
1355             fjx0             = _mm_setzero_ps();
1356             fjy0             = _mm_setzero_ps();
1357             fjz0             = _mm_setzero_ps();
1358             fjx1             = _mm_setzero_ps();
1359             fjy1             = _mm_setzero_ps();
1360             fjz1             = _mm_setzero_ps();
1361             fjx2             = _mm_setzero_ps();
1362             fjy2             = _mm_setzero_ps();
1363             fjz2             = _mm_setzero_ps();
1364             fjx3             = _mm_setzero_ps();
1365             fjy3             = _mm_setzero_ps();
1366             fjz3             = _mm_setzero_ps();
1367
1368             /**************************
1369              * CALCULATE INTERACTIONS *
1370              **************************/
1371
1372             r00              = _mm_mul_ps(rsq00,rinv00);
1373
1374             /* Calculate table index by multiplying r with table scale and truncate to integer */
1375             rt               = _mm_mul_ps(r00,vftabscale);
1376             vfitab           = _mm_cvttps_epi32(rt);
1377             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1378             vfitab           = _mm_slli_epi32(vfitab,3);
1379
1380             /* CUBIC SPLINE TABLE DISPERSION */
1381             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1382             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1383             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1384             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1385             _MM_TRANSPOSE4_PS(Y,F,G,H);
1386             Heps             = _mm_mul_ps(vfeps,H);
1387             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1388             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1389             fvdw6            = _mm_mul_ps(c6_00,FF);
1390
1391             /* CUBIC SPLINE TABLE REPULSION */
1392             vfitab           = _mm_add_epi32(vfitab,ifour);
1393             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1394             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1395             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1396             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1397             _MM_TRANSPOSE4_PS(Y,F,G,H);
1398             Heps             = _mm_mul_ps(vfeps,H);
1399             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1400             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1401             fvdw12           = _mm_mul_ps(c12_00,FF);
1402             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1403
1404             fscal            = fvdw;
1405
1406             /* Calculate temporary vectorial force */
1407             tx               = _mm_mul_ps(fscal,dx00);
1408             ty               = _mm_mul_ps(fscal,dy00);
1409             tz               = _mm_mul_ps(fscal,dz00);
1410
1411             /* Update vectorial force */
1412             fix0             = _mm_add_ps(fix0,tx);
1413             fiy0             = _mm_add_ps(fiy0,ty);
1414             fiz0             = _mm_add_ps(fiz0,tz);
1415
1416             fjx0             = _mm_add_ps(fjx0,tx);
1417             fjy0             = _mm_add_ps(fjy0,ty);
1418             fjz0             = _mm_add_ps(fjz0,tz);
1419
1420             /**************************
1421              * CALCULATE INTERACTIONS *
1422              **************************/
1423
1424             /* COULOMB ELECTROSTATICS */
1425             velec            = _mm_mul_ps(qq11,rinv11);
1426             felec            = _mm_mul_ps(velec,rinvsq11);
1427
1428             fscal            = felec;
1429
1430             /* Calculate temporary vectorial force */
1431             tx               = _mm_mul_ps(fscal,dx11);
1432             ty               = _mm_mul_ps(fscal,dy11);
1433             tz               = _mm_mul_ps(fscal,dz11);
1434
1435             /* Update vectorial force */
1436             fix1             = _mm_add_ps(fix1,tx);
1437             fiy1             = _mm_add_ps(fiy1,ty);
1438             fiz1             = _mm_add_ps(fiz1,tz);
1439
1440             fjx1             = _mm_add_ps(fjx1,tx);
1441             fjy1             = _mm_add_ps(fjy1,ty);
1442             fjz1             = _mm_add_ps(fjz1,tz);
1443
1444             /**************************
1445              * CALCULATE INTERACTIONS *
1446              **************************/
1447
1448             /* COULOMB ELECTROSTATICS */
1449             velec            = _mm_mul_ps(qq12,rinv12);
1450             felec            = _mm_mul_ps(velec,rinvsq12);
1451
1452             fscal            = felec;
1453
1454             /* Calculate temporary vectorial force */
1455             tx               = _mm_mul_ps(fscal,dx12);
1456             ty               = _mm_mul_ps(fscal,dy12);
1457             tz               = _mm_mul_ps(fscal,dz12);
1458
1459             /* Update vectorial force */
1460             fix1             = _mm_add_ps(fix1,tx);
1461             fiy1             = _mm_add_ps(fiy1,ty);
1462             fiz1             = _mm_add_ps(fiz1,tz);
1463
1464             fjx2             = _mm_add_ps(fjx2,tx);
1465             fjy2             = _mm_add_ps(fjy2,ty);
1466             fjz2             = _mm_add_ps(fjz2,tz);
1467
1468             /**************************
1469              * CALCULATE INTERACTIONS *
1470              **************************/
1471
1472             /* COULOMB ELECTROSTATICS */
1473             velec            = _mm_mul_ps(qq13,rinv13);
1474             felec            = _mm_mul_ps(velec,rinvsq13);
1475
1476             fscal            = felec;
1477
1478             /* Calculate temporary vectorial force */
1479             tx               = _mm_mul_ps(fscal,dx13);
1480             ty               = _mm_mul_ps(fscal,dy13);
1481             tz               = _mm_mul_ps(fscal,dz13);
1482
1483             /* Update vectorial force */
1484             fix1             = _mm_add_ps(fix1,tx);
1485             fiy1             = _mm_add_ps(fiy1,ty);
1486             fiz1             = _mm_add_ps(fiz1,tz);
1487
1488             fjx3             = _mm_add_ps(fjx3,tx);
1489             fjy3             = _mm_add_ps(fjy3,ty);
1490             fjz3             = _mm_add_ps(fjz3,tz);
1491
1492             /**************************
1493              * CALCULATE INTERACTIONS *
1494              **************************/
1495
1496             /* COULOMB ELECTROSTATICS */
1497             velec            = _mm_mul_ps(qq21,rinv21);
1498             felec            = _mm_mul_ps(velec,rinvsq21);
1499
1500             fscal            = felec;
1501
1502             /* Calculate temporary vectorial force */
1503             tx               = _mm_mul_ps(fscal,dx21);
1504             ty               = _mm_mul_ps(fscal,dy21);
1505             tz               = _mm_mul_ps(fscal,dz21);
1506
1507             /* Update vectorial force */
1508             fix2             = _mm_add_ps(fix2,tx);
1509             fiy2             = _mm_add_ps(fiy2,ty);
1510             fiz2             = _mm_add_ps(fiz2,tz);
1511
1512             fjx1             = _mm_add_ps(fjx1,tx);
1513             fjy1             = _mm_add_ps(fjy1,ty);
1514             fjz1             = _mm_add_ps(fjz1,tz);
1515
1516             /**************************
1517              * CALCULATE INTERACTIONS *
1518              **************************/
1519
1520             /* COULOMB ELECTROSTATICS */
1521             velec            = _mm_mul_ps(qq22,rinv22);
1522             felec            = _mm_mul_ps(velec,rinvsq22);
1523
1524             fscal            = felec;
1525
1526             /* Calculate temporary vectorial force */
1527             tx               = _mm_mul_ps(fscal,dx22);
1528             ty               = _mm_mul_ps(fscal,dy22);
1529             tz               = _mm_mul_ps(fscal,dz22);
1530
1531             /* Update vectorial force */
1532             fix2             = _mm_add_ps(fix2,tx);
1533             fiy2             = _mm_add_ps(fiy2,ty);
1534             fiz2             = _mm_add_ps(fiz2,tz);
1535
1536             fjx2             = _mm_add_ps(fjx2,tx);
1537             fjy2             = _mm_add_ps(fjy2,ty);
1538             fjz2             = _mm_add_ps(fjz2,tz);
1539
1540             /**************************
1541              * CALCULATE INTERACTIONS *
1542              **************************/
1543
1544             /* COULOMB ELECTROSTATICS */
1545             velec            = _mm_mul_ps(qq23,rinv23);
1546             felec            = _mm_mul_ps(velec,rinvsq23);
1547
1548             fscal            = felec;
1549
1550             /* Calculate temporary vectorial force */
1551             tx               = _mm_mul_ps(fscal,dx23);
1552             ty               = _mm_mul_ps(fscal,dy23);
1553             tz               = _mm_mul_ps(fscal,dz23);
1554
1555             /* Update vectorial force */
1556             fix2             = _mm_add_ps(fix2,tx);
1557             fiy2             = _mm_add_ps(fiy2,ty);
1558             fiz2             = _mm_add_ps(fiz2,tz);
1559
1560             fjx3             = _mm_add_ps(fjx3,tx);
1561             fjy3             = _mm_add_ps(fjy3,ty);
1562             fjz3             = _mm_add_ps(fjz3,tz);
1563
1564             /**************************
1565              * CALCULATE INTERACTIONS *
1566              **************************/
1567
1568             /* COULOMB ELECTROSTATICS */
1569             velec            = _mm_mul_ps(qq31,rinv31);
1570             felec            = _mm_mul_ps(velec,rinvsq31);
1571
1572             fscal            = felec;
1573
1574             /* Calculate temporary vectorial force */
1575             tx               = _mm_mul_ps(fscal,dx31);
1576             ty               = _mm_mul_ps(fscal,dy31);
1577             tz               = _mm_mul_ps(fscal,dz31);
1578
1579             /* Update vectorial force */
1580             fix3             = _mm_add_ps(fix3,tx);
1581             fiy3             = _mm_add_ps(fiy3,ty);
1582             fiz3             = _mm_add_ps(fiz3,tz);
1583
1584             fjx1             = _mm_add_ps(fjx1,tx);
1585             fjy1             = _mm_add_ps(fjy1,ty);
1586             fjz1             = _mm_add_ps(fjz1,tz);
1587
1588             /**************************
1589              * CALCULATE INTERACTIONS *
1590              **************************/
1591
1592             /* COULOMB ELECTROSTATICS */
1593             velec            = _mm_mul_ps(qq32,rinv32);
1594             felec            = _mm_mul_ps(velec,rinvsq32);
1595
1596             fscal            = felec;
1597
1598             /* Calculate temporary vectorial force */
1599             tx               = _mm_mul_ps(fscal,dx32);
1600             ty               = _mm_mul_ps(fscal,dy32);
1601             tz               = _mm_mul_ps(fscal,dz32);
1602
1603             /* Update vectorial force */
1604             fix3             = _mm_add_ps(fix3,tx);
1605             fiy3             = _mm_add_ps(fiy3,ty);
1606             fiz3             = _mm_add_ps(fiz3,tz);
1607
1608             fjx2             = _mm_add_ps(fjx2,tx);
1609             fjy2             = _mm_add_ps(fjy2,ty);
1610             fjz2             = _mm_add_ps(fjz2,tz);
1611
1612             /**************************
1613              * CALCULATE INTERACTIONS *
1614              **************************/
1615
1616             /* COULOMB ELECTROSTATICS */
1617             velec            = _mm_mul_ps(qq33,rinv33);
1618             felec            = _mm_mul_ps(velec,rinvsq33);
1619
1620             fscal            = felec;
1621
1622             /* Calculate temporary vectorial force */
1623             tx               = _mm_mul_ps(fscal,dx33);
1624             ty               = _mm_mul_ps(fscal,dy33);
1625             tz               = _mm_mul_ps(fscal,dz33);
1626
1627             /* Update vectorial force */
1628             fix3             = _mm_add_ps(fix3,tx);
1629             fiy3             = _mm_add_ps(fiy3,ty);
1630             fiz3             = _mm_add_ps(fiz3,tz);
1631
1632             fjx3             = _mm_add_ps(fjx3,tx);
1633             fjy3             = _mm_add_ps(fjy3,ty);
1634             fjz3             = _mm_add_ps(fjz3,tz);
1635
1636             fjptrA             = f+j_coord_offsetA;
1637             fjptrB             = f+j_coord_offsetB;
1638             fjptrC             = f+j_coord_offsetC;
1639             fjptrD             = f+j_coord_offsetD;
1640
1641             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1642                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1643                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1644
1645             /* Inner loop uses 294 flops */
1646         }
1647
1648         if(jidx<j_index_end)
1649         {
1650
1651             /* Get j neighbor index, and coordinate index */
1652             jnrlistA         = jjnr[jidx];
1653             jnrlistB         = jjnr[jidx+1];
1654             jnrlistC         = jjnr[jidx+2];
1655             jnrlistD         = jjnr[jidx+3];
1656             /* Sign of each element will be negative for non-real atoms.
1657              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1658              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1659              */
1660             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1661             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1662             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1663             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1664             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1665             j_coord_offsetA  = DIM*jnrA;
1666             j_coord_offsetB  = DIM*jnrB;
1667             j_coord_offsetC  = DIM*jnrC;
1668             j_coord_offsetD  = DIM*jnrD;
1669
1670             /* load j atom coordinates */
1671             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1672                                               x+j_coord_offsetC,x+j_coord_offsetD,
1673                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1674                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1675
1676             /* Calculate displacement vector */
1677             dx00             = _mm_sub_ps(ix0,jx0);
1678             dy00             = _mm_sub_ps(iy0,jy0);
1679             dz00             = _mm_sub_ps(iz0,jz0);
1680             dx11             = _mm_sub_ps(ix1,jx1);
1681             dy11             = _mm_sub_ps(iy1,jy1);
1682             dz11             = _mm_sub_ps(iz1,jz1);
1683             dx12             = _mm_sub_ps(ix1,jx2);
1684             dy12             = _mm_sub_ps(iy1,jy2);
1685             dz12             = _mm_sub_ps(iz1,jz2);
1686             dx13             = _mm_sub_ps(ix1,jx3);
1687             dy13             = _mm_sub_ps(iy1,jy3);
1688             dz13             = _mm_sub_ps(iz1,jz3);
1689             dx21             = _mm_sub_ps(ix2,jx1);
1690             dy21             = _mm_sub_ps(iy2,jy1);
1691             dz21             = _mm_sub_ps(iz2,jz1);
1692             dx22             = _mm_sub_ps(ix2,jx2);
1693             dy22             = _mm_sub_ps(iy2,jy2);
1694             dz22             = _mm_sub_ps(iz2,jz2);
1695             dx23             = _mm_sub_ps(ix2,jx3);
1696             dy23             = _mm_sub_ps(iy2,jy3);
1697             dz23             = _mm_sub_ps(iz2,jz3);
1698             dx31             = _mm_sub_ps(ix3,jx1);
1699             dy31             = _mm_sub_ps(iy3,jy1);
1700             dz31             = _mm_sub_ps(iz3,jz1);
1701             dx32             = _mm_sub_ps(ix3,jx2);
1702             dy32             = _mm_sub_ps(iy3,jy2);
1703             dz32             = _mm_sub_ps(iz3,jz2);
1704             dx33             = _mm_sub_ps(ix3,jx3);
1705             dy33             = _mm_sub_ps(iy3,jy3);
1706             dz33             = _mm_sub_ps(iz3,jz3);
1707
1708             /* Calculate squared distance and things based on it */
1709             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1710             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1711             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1712             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1713             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1714             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1715             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1716             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1717             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1718             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1719
1720             rinv00           = gmx_mm_invsqrt_ps(rsq00);
1721             rinv11           = gmx_mm_invsqrt_ps(rsq11);
1722             rinv12           = gmx_mm_invsqrt_ps(rsq12);
1723             rinv13           = gmx_mm_invsqrt_ps(rsq13);
1724             rinv21           = gmx_mm_invsqrt_ps(rsq21);
1725             rinv22           = gmx_mm_invsqrt_ps(rsq22);
1726             rinv23           = gmx_mm_invsqrt_ps(rsq23);
1727             rinv31           = gmx_mm_invsqrt_ps(rsq31);
1728             rinv32           = gmx_mm_invsqrt_ps(rsq32);
1729             rinv33           = gmx_mm_invsqrt_ps(rsq33);
1730
1731             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
1732             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
1733             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
1734             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
1735             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
1736             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
1737             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
1738             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
1739             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
1740
1741             fjx0             = _mm_setzero_ps();
1742             fjy0             = _mm_setzero_ps();
1743             fjz0             = _mm_setzero_ps();
1744             fjx1             = _mm_setzero_ps();
1745             fjy1             = _mm_setzero_ps();
1746             fjz1             = _mm_setzero_ps();
1747             fjx2             = _mm_setzero_ps();
1748             fjy2             = _mm_setzero_ps();
1749             fjz2             = _mm_setzero_ps();
1750             fjx3             = _mm_setzero_ps();
1751             fjy3             = _mm_setzero_ps();
1752             fjz3             = _mm_setzero_ps();
1753
1754             /**************************
1755              * CALCULATE INTERACTIONS *
1756              **************************/
1757
1758             r00              = _mm_mul_ps(rsq00,rinv00);
1759             r00              = _mm_andnot_ps(dummy_mask,r00);
1760
1761             /* Calculate table index by multiplying r with table scale and truncate to integer */
1762             rt               = _mm_mul_ps(r00,vftabscale);
1763             vfitab           = _mm_cvttps_epi32(rt);
1764             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1765             vfitab           = _mm_slli_epi32(vfitab,3);
1766
1767             /* CUBIC SPLINE TABLE DISPERSION */
1768             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1769             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1770             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1771             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1772             _MM_TRANSPOSE4_PS(Y,F,G,H);
1773             Heps             = _mm_mul_ps(vfeps,H);
1774             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1775             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1776             fvdw6            = _mm_mul_ps(c6_00,FF);
1777
1778             /* CUBIC SPLINE TABLE REPULSION */
1779             vfitab           = _mm_add_epi32(vfitab,ifour);
1780             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1781             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1782             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1783             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1784             _MM_TRANSPOSE4_PS(Y,F,G,H);
1785             Heps             = _mm_mul_ps(vfeps,H);
1786             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1787             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1788             fvdw12           = _mm_mul_ps(c12_00,FF);
1789             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1790
1791             fscal            = fvdw;
1792
1793             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1794
1795             /* Calculate temporary vectorial force */
1796             tx               = _mm_mul_ps(fscal,dx00);
1797             ty               = _mm_mul_ps(fscal,dy00);
1798             tz               = _mm_mul_ps(fscal,dz00);
1799
1800             /* Update vectorial force */
1801             fix0             = _mm_add_ps(fix0,tx);
1802             fiy0             = _mm_add_ps(fiy0,ty);
1803             fiz0             = _mm_add_ps(fiz0,tz);
1804
1805             fjx0             = _mm_add_ps(fjx0,tx);
1806             fjy0             = _mm_add_ps(fjy0,ty);
1807             fjz0             = _mm_add_ps(fjz0,tz);
1808
1809             /**************************
1810              * CALCULATE INTERACTIONS *
1811              **************************/
1812
1813             /* COULOMB ELECTROSTATICS */
1814             velec            = _mm_mul_ps(qq11,rinv11);
1815             felec            = _mm_mul_ps(velec,rinvsq11);
1816
1817             fscal            = felec;
1818
1819             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1820
1821             /* Calculate temporary vectorial force */
1822             tx               = _mm_mul_ps(fscal,dx11);
1823             ty               = _mm_mul_ps(fscal,dy11);
1824             tz               = _mm_mul_ps(fscal,dz11);
1825
1826             /* Update vectorial force */
1827             fix1             = _mm_add_ps(fix1,tx);
1828             fiy1             = _mm_add_ps(fiy1,ty);
1829             fiz1             = _mm_add_ps(fiz1,tz);
1830
1831             fjx1             = _mm_add_ps(fjx1,tx);
1832             fjy1             = _mm_add_ps(fjy1,ty);
1833             fjz1             = _mm_add_ps(fjz1,tz);
1834
1835             /**************************
1836              * CALCULATE INTERACTIONS *
1837              **************************/
1838
1839             /* COULOMB ELECTROSTATICS */
1840             velec            = _mm_mul_ps(qq12,rinv12);
1841             felec            = _mm_mul_ps(velec,rinvsq12);
1842
1843             fscal            = felec;
1844
1845             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1846
1847             /* Calculate temporary vectorial force */
1848             tx               = _mm_mul_ps(fscal,dx12);
1849             ty               = _mm_mul_ps(fscal,dy12);
1850             tz               = _mm_mul_ps(fscal,dz12);
1851
1852             /* Update vectorial force */
1853             fix1             = _mm_add_ps(fix1,tx);
1854             fiy1             = _mm_add_ps(fiy1,ty);
1855             fiz1             = _mm_add_ps(fiz1,tz);
1856
1857             fjx2             = _mm_add_ps(fjx2,tx);
1858             fjy2             = _mm_add_ps(fjy2,ty);
1859             fjz2             = _mm_add_ps(fjz2,tz);
1860
1861             /**************************
1862              * CALCULATE INTERACTIONS *
1863              **************************/
1864
1865             /* COULOMB ELECTROSTATICS */
1866             velec            = _mm_mul_ps(qq13,rinv13);
1867             felec            = _mm_mul_ps(velec,rinvsq13);
1868
1869             fscal            = felec;
1870
1871             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1872
1873             /* Calculate temporary vectorial force */
1874             tx               = _mm_mul_ps(fscal,dx13);
1875             ty               = _mm_mul_ps(fscal,dy13);
1876             tz               = _mm_mul_ps(fscal,dz13);
1877
1878             /* Update vectorial force */
1879             fix1             = _mm_add_ps(fix1,tx);
1880             fiy1             = _mm_add_ps(fiy1,ty);
1881             fiz1             = _mm_add_ps(fiz1,tz);
1882
1883             fjx3             = _mm_add_ps(fjx3,tx);
1884             fjy3             = _mm_add_ps(fjy3,ty);
1885             fjz3             = _mm_add_ps(fjz3,tz);
1886
1887             /**************************
1888              * CALCULATE INTERACTIONS *
1889              **************************/
1890
1891             /* COULOMB ELECTROSTATICS */
1892             velec            = _mm_mul_ps(qq21,rinv21);
1893             felec            = _mm_mul_ps(velec,rinvsq21);
1894
1895             fscal            = felec;
1896
1897             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1898
1899             /* Calculate temporary vectorial force */
1900             tx               = _mm_mul_ps(fscal,dx21);
1901             ty               = _mm_mul_ps(fscal,dy21);
1902             tz               = _mm_mul_ps(fscal,dz21);
1903
1904             /* Update vectorial force */
1905             fix2             = _mm_add_ps(fix2,tx);
1906             fiy2             = _mm_add_ps(fiy2,ty);
1907             fiz2             = _mm_add_ps(fiz2,tz);
1908
1909             fjx1             = _mm_add_ps(fjx1,tx);
1910             fjy1             = _mm_add_ps(fjy1,ty);
1911             fjz1             = _mm_add_ps(fjz1,tz);
1912
1913             /**************************
1914              * CALCULATE INTERACTIONS *
1915              **************************/
1916
1917             /* COULOMB ELECTROSTATICS */
1918             velec            = _mm_mul_ps(qq22,rinv22);
1919             felec            = _mm_mul_ps(velec,rinvsq22);
1920
1921             fscal            = felec;
1922
1923             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1924
1925             /* Calculate temporary vectorial force */
1926             tx               = _mm_mul_ps(fscal,dx22);
1927             ty               = _mm_mul_ps(fscal,dy22);
1928             tz               = _mm_mul_ps(fscal,dz22);
1929
1930             /* Update vectorial force */
1931             fix2             = _mm_add_ps(fix2,tx);
1932             fiy2             = _mm_add_ps(fiy2,ty);
1933             fiz2             = _mm_add_ps(fiz2,tz);
1934
1935             fjx2             = _mm_add_ps(fjx2,tx);
1936             fjy2             = _mm_add_ps(fjy2,ty);
1937             fjz2             = _mm_add_ps(fjz2,tz);
1938
1939             /**************************
1940              * CALCULATE INTERACTIONS *
1941              **************************/
1942
1943             /* COULOMB ELECTROSTATICS */
1944             velec            = _mm_mul_ps(qq23,rinv23);
1945             felec            = _mm_mul_ps(velec,rinvsq23);
1946
1947             fscal            = felec;
1948
1949             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1950
1951             /* Calculate temporary vectorial force */
1952             tx               = _mm_mul_ps(fscal,dx23);
1953             ty               = _mm_mul_ps(fscal,dy23);
1954             tz               = _mm_mul_ps(fscal,dz23);
1955
1956             /* Update vectorial force */
1957             fix2             = _mm_add_ps(fix2,tx);
1958             fiy2             = _mm_add_ps(fiy2,ty);
1959             fiz2             = _mm_add_ps(fiz2,tz);
1960
1961             fjx3             = _mm_add_ps(fjx3,tx);
1962             fjy3             = _mm_add_ps(fjy3,ty);
1963             fjz3             = _mm_add_ps(fjz3,tz);
1964
1965             /**************************
1966              * CALCULATE INTERACTIONS *
1967              **************************/
1968
1969             /* COULOMB ELECTROSTATICS */
1970             velec            = _mm_mul_ps(qq31,rinv31);
1971             felec            = _mm_mul_ps(velec,rinvsq31);
1972
1973             fscal            = felec;
1974
1975             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1976
1977             /* Calculate temporary vectorial force */
1978             tx               = _mm_mul_ps(fscal,dx31);
1979             ty               = _mm_mul_ps(fscal,dy31);
1980             tz               = _mm_mul_ps(fscal,dz31);
1981
1982             /* Update vectorial force */
1983             fix3             = _mm_add_ps(fix3,tx);
1984             fiy3             = _mm_add_ps(fiy3,ty);
1985             fiz3             = _mm_add_ps(fiz3,tz);
1986
1987             fjx1             = _mm_add_ps(fjx1,tx);
1988             fjy1             = _mm_add_ps(fjy1,ty);
1989             fjz1             = _mm_add_ps(fjz1,tz);
1990
1991             /**************************
1992              * CALCULATE INTERACTIONS *
1993              **************************/
1994
1995             /* COULOMB ELECTROSTATICS */
1996             velec            = _mm_mul_ps(qq32,rinv32);
1997             felec            = _mm_mul_ps(velec,rinvsq32);
1998
1999             fscal            = felec;
2000
2001             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2002
2003             /* Calculate temporary vectorial force */
2004             tx               = _mm_mul_ps(fscal,dx32);
2005             ty               = _mm_mul_ps(fscal,dy32);
2006             tz               = _mm_mul_ps(fscal,dz32);
2007
2008             /* Update vectorial force */
2009             fix3             = _mm_add_ps(fix3,tx);
2010             fiy3             = _mm_add_ps(fiy3,ty);
2011             fiz3             = _mm_add_ps(fiz3,tz);
2012
2013             fjx2             = _mm_add_ps(fjx2,tx);
2014             fjy2             = _mm_add_ps(fjy2,ty);
2015             fjz2             = _mm_add_ps(fjz2,tz);
2016
2017             /**************************
2018              * CALCULATE INTERACTIONS *
2019              **************************/
2020
2021             /* COULOMB ELECTROSTATICS */
2022             velec            = _mm_mul_ps(qq33,rinv33);
2023             felec            = _mm_mul_ps(velec,rinvsq33);
2024
2025             fscal            = felec;
2026
2027             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2028
2029             /* Calculate temporary vectorial force */
2030             tx               = _mm_mul_ps(fscal,dx33);
2031             ty               = _mm_mul_ps(fscal,dy33);
2032             tz               = _mm_mul_ps(fscal,dz33);
2033
2034             /* Update vectorial force */
2035             fix3             = _mm_add_ps(fix3,tx);
2036             fiy3             = _mm_add_ps(fiy3,ty);
2037             fiz3             = _mm_add_ps(fiz3,tz);
2038
2039             fjx3             = _mm_add_ps(fjx3,tx);
2040             fjy3             = _mm_add_ps(fjy3,ty);
2041             fjz3             = _mm_add_ps(fjz3,tz);
2042
2043             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2044             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2045             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2046             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2047
2048             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
2049                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2050                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2051
2052             /* Inner loop uses 295 flops */
2053         }
2054
2055         /* End of innermost loop */
2056
2057         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2058                                               f+i_coord_offset,fshift+i_shift_offset);
2059
2060         /* Increment number of inner iterations */
2061         inneriter                  += j_index_end - j_index_start;
2062
2063         /* Outer loop uses 24 flops */
2064     }
2065
2066     /* Increment number of outer iterations */
2067     outeriter        += nri;
2068
2069     /* Update outer/inner flops */
2070
2071     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*295);
2072 }